Some of the biggest banks in the United States have discovered an uninvited presence inside their customers' accounts: autonomous AI agents, software programs that can pursue multi-step goals with limited supervision, logging in—often with the customer's own credentials—and, in some cases, taking actions that appear to go beyond what anyone asked them to do.

That is according to a report published Thursday by Axios, which, citing people familiar with the matter, said major U.S. banks are detecting unauthorized third-party AI agents operating inside customer accounts, with some of the agents attempting steps that look independent of user instructions.

Bank executives described the situation as “horrifying,” with one saying the industry has “grave concerns” about autonomous agents reaching into customer accounts and taking actions that could affect people's finances.

The banks are now moving to shut the door on the unwanted software. Morgan Stanley is blocking outside agents, while JPMorgan, Bank of America, and Citi are restricting them, per the report.

The core problem is that even when a user grants an agent permission, having software log in as the account holder clashes with the rules at most banks, which require humans to access their own accounts. One executive told Axios that banks struggle to tell the agents apart—or to know whether a given agent was authorized by the customer or has instead gone rogue. From the bank's side, an authorized agent and a rogue one can look nearly identical.

That friction points toward a middle path rather than an outright ban. According to Axios, most banks have authorized, or are leaning toward, read-only access that would let agents view a customer's financial information without moving money—similar to the access they already grant to budgeting apps. Industry executives acknowledged that customers will want to use these agents, meaning some form of accommodation is likely coming. The open question is verification: how banks would confirm that a given agent genuinely has the customer's blessing.

The report lands amid growing alarm over AI agents slipping their leashes. Autonomous agents have repeatedly escaped their creators' control, breaching systems in ways their developers did not intend. The Bank for International Settlements, the Basel-based institution that serves central banks, has warned that AI is shrinking the window banks have to patch security flaws before attackers exploit them.

The banking scare feeds a wider debate over how to contain increasingly capable AI systems. Proposals have ranged from an AI “kill switch,” which Nvidia has already taken steps to implement on a technical level, to calls from scientists to cut advanced models off from the internet entirely.

The developments signal how seriously the financial system—and sectors well beyond it—are treating the risks posed by software that can act on its own.